reflectance tranformation imaging | RTI

Post 1: RTI Recording Session: 11 November 2015

1. Pre recording/Introduction:

10th November 2015: As a group we held a discussion about the set of objects we were recording, what features we could capture and how and which items to investigate more closely.See table, Fig. 1,below:

1-Table proj mgmt_1
Fig. 1_Artefact + recording method table.

We decided to use Reflectance Transformation Imaging (RTI) on the Greek coin, the coarsely textured papyrus and the canvas painting.

‘RTI is a technique that uses digital cameras and lights to capture and enhance subtle surface details, and generate detailed surface models of objects and their interactive re-lighting’ ( quoted in Moodle/AFF622_WeekSixPresentation_2015-16.pdf)

The first algorithm for photo amalgamation was developed by Tom Malzbender of HP labs in 2001, it is referred to as Polynomial Texture Mapping (PTM).All three artefacts have surface textures that are not very obvious when viewed in a photograph. In the case of the coin, the surface variations aren’t very visible due to its small size (24mm) whereas with the papyrus and painting, their integral and surface textures are not significantly dramatic enough to show themselves in a photograph taken from a standard front view. We itemised the equipment we would need to successfully capture these textures and decided to use the image laboratory to take our photographs as it is possible to control the lighting almost completely there.

2. Day 1; image capture, 11th November:

Firstly, we uploaded the Canon EOS software on to our laptops and set up a destination folder for the images and the image types we wanted to save (RAW and jpeg). As we were photographing the coin first, we decided to use a 100mm macro lens attached to the inverted tripod (camera pointing lens downward).

3. Greek coin specifications; side one (Fig. 2):

Fig. 2. Greek coin circa 1960, side one with shiny spheres, grey card and ruler (after it moved).

      a. 24mm in diameter
      b. 1.2mm thick approximately.
      c. Made from a light metal or alloy, possibly aluminium.
      d. Distinctive features; hole in the centre, chip on the side, slight warp in profile.

More information would be recorded in a cultural heritage setting (such as weight, composition, graphical elements, age and provenance) as part of the accompanying digital metadata record of an object but as this is a 2.5D recording process, we were recording visible and textural information. The RTI tool is a method that exaggerates textural features for digital viewing.

4. Set up and test:

We set up our camera and tripod over the target object (coin), and set everything else up to avoid moving the target object and to facilitate a full 360° access to it on all sides. The coin was placed on a dark grey Kappa board(taped to the floor), with two 2.5mm ball bearings acting as spheres, a grey (18%)colour balance card and a ruler to act as a scale. We chose the 2.5 mm size of ball bearing because it was the closest in thickness to the object without being microscopic and it didn’t cast shadows over the coin when the torch was held at a low angle. Next we set the lens on auto focus and took some test shots(Fig. 3):

Fig. 3. Camera test table. Greek coin, side one.

After selecting *f11 and an ISO(the measure of a photographic film’s sensitivity to light(Wikipedia)) of 100 we switched to manual focus (without removing the torch from lighting the object) and did another test shot to confirm that we had the best settings for the recording. We began recording , holding the LED torch at 15°, 35° and 60° angles to the centre of the object and using a string to measure the preferred(according to CHI guidelines) distance to hold the torch from it (diameter x 3).
A small object like the coin is most likely to get moved because of its tiny size and the person with the torch accidentally touching it whilst making a measurement with the string.
This happened during this set of captures; the ruler accidentally moved. We stopped photographing and discussed whether we needed to start from the beginning again. We decided that because it was not the base-board, spheres or actual coin that moved we could carry on and finish taking the image set because the important elements above were still unmoved and in their place. Sometimes the co-ordination between the person working on the laptop and the person with the torch would get confused and a photograph would get taken at the wrong time. These image numbers were noted, and removed at the post processing stage in RTI Builder. The image capturing stage took about seventeen minutes and we took thirty eight images.

5. Post-Processing:

Before opening RTI Builder we set up the folder directory structure for the captured RTI files, see below(Fig. 4a):


Fig. 4a. Folder hierarchy for RTI Builder_before processing

Coin in RTI_builder
Fig. 4b. RTI Builder interface, centering spheres.

When we opened RTI Builder we:

      1.Removed any of the noted ‘bad’ images from the image set.
      2.Selected the red spheres option and selected the shiny spheres.
      3.Centered spheres within selection.
      4.Highlight detection (tracks the movement of light over the recorded object over the span of the image set).
      5.Cropped ruler, grey card and spheres from the image set whilst centering our object in the final image.

At the end of this process we had more folders added to the directory structure(Fig.5):


Fig. 4. Folder hierarchy for RTI Builder_after processing

6. PTM Fitter + RTI – Viewer:

We found the pre- downloaded (from Hewlett Packard) software for the PTM Fitter on our laptop and ran the software on the ‘built’ image file. Using the RTI Viewer (version 1.1) we could mouse-over the sphere in the top right menu to see the effect of what directional light could look like cast over the object in the dark. Different filter headings give different effects i.e. Multispectral light (exaggerated shine). We varied our choices here from object to object as one type was not the best ‘fit’ for every object.

7. Greek coin: Side Two

We turned the coin over to capture its other side and began the process of refocusing the image and testing f-stop/ISO/exposure times again. When this sort of preparatory work was done the image capture was quickly done. The RTI Build and RTI Viewer stages were also quicker to complete this time (27 minutes)

8. Papyrus(Fig.6) specifications and capture:

425mm Length x 345mm Width.
(Please note that these measurements are approximate only due to the uneven and variable nature of the edges of the object)

Fig. 6. Egyptian Papyrus. Painted reproduction.

This is a much larger object to capture than the Greek coin so we changed from macro to a normal camera lens and raised the height of the camera in the centre of the tripod. We were drawn towards just capturing interesting details of it as the ragged sides and surface painting detail are interesting but with no specific details that couldn’t be seen in a normal photograph. Capturing the whole of it would allow us to see parts of it in magnification to examine details,as well as having the whole, so we decided to capture the whole of it. Also, the surface shape and texture of papyrus is unique to this type of material(a woven lattice of vegetable pulp with long fibres, quite unlike the paper we use today).
After testing and focussing the camera, the settings we decided on were f11, ISO 2000 and 1/13 of a second exposure time. A high ISO was needed because our torch was just able to light the whole papyrus in every shot, but not with a great deal of strength. This set of images was complicated by the fact that the material is fragile, the side edges curl and it doesn’t sit flat on any surface in spite of spending over 24 hours of being persuaded to flatten between two panes of glass. In a museum setting, if a vacuum table (to hold materials flat using suction) was available this would be a good object to use it with and the expertise of a conservator could be called upon for advice on the least destructive technique for image capture. In our situation, we managed to keep the edges from springing back into a roll and took our pictures quickly. We had more images to delete from the image set this time but had a complete set of images to process successfully.

9. Discussion:

We discussed using the RTI process on all of the objects given to us as most of them have interesting texture, even with the ceramic figurines there is a difference in the materials used between the bull figurine and the abstract head. However the differences between these textures are perhaps not unique or interesting enough to warrant capturing as the difference between a smooth ceramic and a slightly textured one is something that can be captured in a standard photograph.

10. Painting specifications and capture:

300mm Length x 237mm Width x 15mm Depth.
With the painting, there are textural marks made on the paint surface that are subtle but interesting. We decided to capture them using RTI to see if we could enhance and identify this sort of texture. This object has a shallow grain from the canvas as well as the paint applied to it. In John’s blog post here about Hyperspectral scanning, and one of my next blogs (see next post below) we will discuss what we found by examining the responses of different paints/materials to different light frequencies. In this case, to record the surface texture with RTI we needed to choose other sizes of spheres and re-test our focus and image resolution in a series of test images as shown below (Fig. 7):

5-painting specs
Fig. 7. Camera test table for Abstract painting (artefact #1).

We had a debate about how best to photograph the object to exaggerate the relatively subtle texture variation on its surface. It was suggested that we take more low angled photos to do this as the higher angles 45°/65° seemed to ‘blank-out’ the texture as with a standard front-view photograph. There was concern that at lower angles the spheres could cast shadows on the object but we decided to try it as only the top third of the sphere needed to be in shot and we wanted to maximise the texture captured(our new capturing angles were 15°,35°and 50°).
After processing in RTI Builder and viewing in RTI Viewer the painting didn’t seem to be yielding anything special. However when viewed using the Diffuse Gain filter we could see that the marks on the canvas were clearly made by pressing fingers flat onto the paint, something that we couldn’t have seen as easily with the naked eye(Fig. 8a + 8b).

CUpaintTextureRTI viewer_animate2

CUpaintTextureRTI viewer_animate1

Fig. 8a + 8b_Textured finger prints in paint surface of Abstract Painting visualised using Diffuse Gain filter in RTI Viewer.

This filter exaggerated the paint texture and shadows as shown in this slide from Piquette and Crowther (31)( Fig. 9):

Filters explained diffuse gain
Fig. 9 RTI Viewer’s filter effects visualised.

11. Critical evaluation of RTI process:

As a group we evaluated our methodology and at times whether to continue the way that we started our image capture as situations arose. The image capture for RTI was relatively trouble free, with occasional pauses and re-evaluations along the way. The way that we could process our images almost immediately on site in Builder and Viewer allowed us to self-assess without losing a lot of time unlike some of the other processes that we used for other objects. As a result, the RTI capture and processing was complete within one day. The quick points of our critical reflection on the RTI process as outlined in our final presentation were;

      a. The camera, object and tripod set-up had to be planned and thought out in advance.
      b. The capturing process was quick but had to be done carefully.
      c. If some part of the set-up did get moved, the re-shoot had to be meticulous and focused.
      d. If our objects had been any bigger we would have had to use a flash (with UV filter) combined with a radio controlled trigger/receiver.
      e. The larger the object the more empty space needed around it to allow lighting distance (The diagonal measurement of the artefact multiplied by two or three)
      f. The RTI Viewer, Builder and webViewer software were not very intuitive and in some cases not easy for a beginner to use.

        g. If I were to do this project again. I would make sure to photograph the Greek coin centered and the right way up.

Lastly Richard was successful in his attempted to embed an RTI Viewer showing one of the finished RTI files within his blog and John has added highlights and annotations to our files in RTI Viewer.

    “Reflectance Transformation Imaging. Guide to Highlight Image Capture” 2015. Cultural Heritage Imaging. version 2.0. Web 10 Nov. 2015

    “Reflectance Transformation Imaging. Guide to RTI Viewer” 2015. Cultural Heritage Imaging. version 2.0.
    Web 10 Nov. 2015

    “Reflectance Transformation Imaging. Guide to RTI Viewer” 2015. Cultural Heritage Imaging. version 2.0. Document version 1.4 for RTIBuilder v. 2.0.2beta (draft version)

    Palma G. “Reflectance Transformation Imaging” Web 24 November 2015.

    Piquette, K. E., Crowther, C. 2011 “Developing a Reflectance Transformation Imaging (RTI) System for Inscription Documentation in Museum Collections and the Field. Case studies on ancient Egyptian and Classical material”. Web 27 Nov. 2015.

    Papadopoulos, K. 2015. AFF622_WeekSixPresentation_2015-16.pdf (7) Web 10 November 2015.


Post 2:Image Capture for Photogrammetry: 16-18 November 2015

1. Pre recording/Introduction:

1. Pre recording/ Introduction:
Following on from Richards post on photogrammetry
this post is going to describe and analyse some of the successes and challenges we worked through whilst capturing and processing our image set for the creation of 3d models from our small figurine/ artefacts. We decided to capture the artefacts; given names here for the purposes of identification, as; ‘the bull’, (or artefact #4) ‘stone head’ (or artefact #5) and ‘water bearer’ (or artefact # 6) using photogrammetry software (Photoscan). Our end goal was to publish the digital constructions produced, in two forms. First; by publishing visualisations of our 3d models (within our blogs) using Sketchfab (
Secondly, our goal was to 3d print our 3d models, through the Maynooth University library 3D printing facility (

2. A rocky start, 16 to 17th November:

On our first attempt at setting up the imaging lab for image capturing we encountered two challenges: The first challenge was choosing a suitable light box. A light box is used in object photography to distribute light evenly over the object’s surface and to minimise the amount of background visual distraction (‘noise’)and shadows in the images taken. The light box Aveen had built was of a suitable size but the paper walls were not made from thin paper (80gsm) but from a weight of paper (100-120gsm) that didn’t allow light into the interior of the light box as easily as the larger wire framed light box (fabric) already in the imaging lab. Another unforeseen factor in the unsuitability of this homemade light box was that the corners and points where sides of the box met; whilst being essential structurally to keep the handmade light box together, blocked light from its interior because of their thickness (20mm) on each side. As a group we agreed that the bigger light box was over sized for the purpose but preferable to the hand made light box in functional terms.
Our second difficulty was in finding a suitable light bulb to replace the one we had broken during set-up. The bulb was of a very specialist nature: 36Watt, 5500Kelvin (cool/daylight colour temperature) with an ‘Edison’ screw fixing (E27). We checked specialist camera shops in Maynooth, Dublin and online, and then with three suppliers to a lighting specialist and lastly hardware shops but none had a bulb to hand that we could source within a short time frame(2 to 3 days). We decided to try the closest match that we could find in Maynooth, and consider post processing our images in Photoshop (in a batch process) to ‘correct’ the colour of our images. This new bulb, (32W, 3200K, E27) when it was installed, gave a weak light with a very orange colour temperature. It was decided unanimously, to be unsuitable, but our need for more light remained. After testing the additional light from various torches (LED torch and a climbers head torch ) the use of two desk lamps were found to be practicable; one throwing light into the light box at the front and the other, from above at the back. We were ready to set up and test.

3. Set up and test:

After the trials of the days before the set up and testing went relatively smoothly, with one caveat, the oasis we had for propping and turning the artefact around was fine for photographs where the artefact (‘the bull’) was on its ‘feet’ but when it was on its side, parts were overhanging the edges of the oasis and there was a fear that it could fall off. The oasis was swapped for a piece of dark grey foam and photographs were tested and taken.
After this initial capture with Richard and John, Aveen photographed the next two artefacts, the ‘stone head’ (Cycladic head) and ‘water bearer’. At this stage we worked individually in order to cover more tasks. The foam square, even when it was stuck down was not smooth and dragged as the head was turned on its central axis. At this point, Richard, who was masking the ‘the bull’ figure in Photoscan, let us know that we needed to change the surface under the ‘stone head’.He found, in the post-processing stage (Photoscan),the grey foam was difficult to select, it gave a jagged edge to the mask around the artefact. A square of smooth black Kappa board was chosen as the base surface for image capture of the ‘stone head’ (Fig. (i)). It was also chosen because of its obvious colour contrast to the object, which would make it easier to select and mask.

head in lightbox

Fig. (i) Set-up with ‘stone head’ on a black Kappa board base.

The camera was set up at F32 ISO 100 (AV priority setting) but given the difference in light reflection and the size of object, a new test for best focus and depth of field was done as outlined below (Fig. (ii)):


Fig. (ii) Camera test table for ‘stone head’ image capture.

After selecting *f16 and an ISO of 100 Aveen switched to manual focus and began the recording and rotation the object by small degrees for a 60% minimum image overlap. After completing a 360 degree rotation and image capture, the photos were checked and were not in focus. With the Canon EOS camera, every photograph needed to be refocussed before being taken (the same as a half button press on the camera itself). With this lesson learned the following of the image capture process went smoothly but not without one more problem (Fig. (iii)): During the image capture, the lights in the imaging lab came on but weren’t noticed.

Aveen_oh dear
Fig. (iii) Realisation that a quarter of images captured were ‘blue’ like Fig (iv)

The result of this was that a quarter of our image set had a colder colour temperature than the main body of images that captured the warm colour temperature of the original object (Fig. (iv) and (v)). The group discussed whether to re-take these photos and decided that we didn’t have the extra time needed to do so.

stoneH_lights on

stoneH_lights off

Fig (iv) and (v): Left: ‘Blue’ colour captured in image taken with imaging lab lights on and
Right: ‘True colour of object’ captured in image taken with imaging lab lights off.

4. ‘Water bearer’ image capture:

The focus testing and capture process above was repeated here with new settings for this artefact. No challenges were encountered at this or the post –processing stages(Fig. (vi)).


Fig. (vi): ‘Water bearer’ figurine.

5. Critical evaluation of image capturing process for photogrammetry:
    • More light sources were needed than we planned for.
    • Broken light bulb issue delayed image capture.
    • The base that objects were photographed on needed to be smooth and not add to workload at the masking stage.
    • The surface as well as background to the photographed object is very important: The black Kappa board was ideal for the ‘stone head’. In retrospect, a white surface would have been better for all of the artefacts, for light reflection and to keep the original colour of the artefact accurate.
    • If we had more time, we could have used colour accuracy cards to ensure that our colours were as accurate to the original artefact as possible. The photographs could have been tested and colour-matched in a software such as Photoshop.
    • The set-up, test photo and masking stages were time consuming.
    • Our 3d models were not saved the right way up (in the Y plane) in Photoscan but can be orientated correctly after upload to Sketchfab.
    • We had planning issues and were not prepared for everything that went wrong.
    • Lighting in the imaging laboratory turned itself on during image capture.

In spite of our difficulties with the image capture we were successful in processing and publishing our 3d models in Sketchfab:
‘The bull’
‘stone head’
‘water bearer’

We were also successful in having a version of our set of artefacts printed in Maynooth University’s 3D printer, see Fig. (vii-ix).




Fig. (vii-ix): 3d printed models made from PLA material: ‘stone head’ , ‘the bull’ and the injured ‘water bearer’.


Post 3: Hyperspectral scanning of Abstract Painting: Colour swatch card


Fig.a: Abstract Painting or artefact #1, shown in unmodified lighting.

1. Michelangelo:


Fig b: Screenshot of a Hyperspectral scan of the drawing under the visible surface of the Abstract Painting.

During Johns scanning of artefact # 1 (Abstract painting) he found a relatively clear
piece of writing that runs as ‘I saw the angel in the marble […] I set him
free’(see John’s post here). Underneath this writing was a vague impression, visible in the red-
orange part of the spectrum (651nanometers), with what looked like the ending ‘elo’ at the end (Fig. c).


Fig. c: Enhanced scan of possible signature ‘Michelangelo’.

In response to the possibility of this find; I produced a colour swatch card (Fig. d) and its key (Fig. e) using a mixture of media, some of which could possibly match the media used in the painting by comparison. If we could match the response of the colours on the swatch with those of the materials used on the canvas this could narrow down what media were possibly used. Perhaps, we might even find a match that could provoke further investigation.

colour card scan_resized
Fig. d: Colour swatch with mixed media.

colour card_keycropd
Fig. e: Key to colour swatch media types.

The outline objectives for the production of the colour swatch were:

    • To see if any of the media on the swatch match those of the writing/ possible signature hidden under the quote.
    • To show by comparison if the paint type on the swatch (acrylic) was a possible match to the paint used on the canvas.
    • To demonstrate how red and yellow are transparent colours and if text was to be concealed underneath either it would need a layer of other more densely pigmented paint over it in order to conceal any writing/drawing.
    • To show the order in which the layers of paint/media were applied to the canvas.
    • To see if the blue/green paint would match the colour of the paint on the canvas when over painted on top of cadmium red, and to see if we could get a similar fingerprint effect when a finger was pressed into the paint.

2. Conclusions:

The cadmium yellow acrylic appeared to respond similarly in terms of light
absorption, when one pixel from the artefact yellow was compared with a pixel from the cadmium yellow on the swatch card (Fig. f).


Fig. f: Screenshot of cadmium yellow comparison between swatch and painting (colours not accurately shown).

We didn’t find a conclusive match to our ‘Michelangelo’ signature in any of the writing media (a-i) on the swatch card. The red and yellow acrylic paints on the swatch card show the transparency of these particular colours in acrylic paint as the pencil written under them, on the swatch, is not concealed by the paint. The titanium white paint however (item q, Fig. d) completely conceals writing in black permanent marker underneath it. The titanium white on the swatch (as with the yellow comparison) seems to be speculatively close (see graph Fig. g)to the white paint on the painting at the time of this screenshot (Fig. g).
At this point I can acknowledge that, without first hand knowledge of the process, in this instance, all that I know that I don’t know. Equally, with our cadmium yellow paint comparisons, I cannot say that this line of enquiry came to any definitive conclusions. It does show how a case could be made perhaps for the use of Hyperspectal scanning in other colour media comparisons as well as for the uncovering of hidden messages under the paint.

match titanWhite_screenshot

Fig. g: Comparisons of titanium white versus the white paint used on Abstract Painting.

Leave a Reply

Your email address will not be published. Required fields are marked *